Security
Controls that do not depend on anyone remembering.
Separation of duties, dual control and an unalterable record of who did what, all enforced in the call path rather than described in a policy document. What follows is specific enough to test, which is the only kind of security claim worth printing.
Four things the system will not let you do.
Controls that depend on somebody remembering are controls that fail quietly. These are in the code path.
Dual control is in the framework, not the policy
Maker-checker is enforced by the authorisation layer every module calls into. A rule that lives in a policy document is a rule somebody can be talked out of at four on a Friday.
Approving is a separate permission from doing
Otherwise two people can be two holders of the same role, and four eyes quietly becomes two, invisibly, because the system still sees two distinct users. A segregation matrix prevents it and a test asserts no role holds both sides.
The maker cannot approve their own work
Rejected by the framework rather than caught in review. The check does not depend on anybody noticing.
Every state change is evidence
Hash-chained, append-only, with the actor and the before value. The database has no update path on it, so an incident investigation starts from a record rather than from a reconstruction.
Authentication and access.
Hosting and operations.
Corlayr is deployed by the institution that runs it. These are properties of what we ship, not of a service we operate on your behalf.
This website.
A static site on Vercel. No third-party analytics, no advertising pixels, no session recording, no chat widget loading somebody else’s JavaScript onto a page where a compliance officer is typing their name.
The demo form posts to our own endpoint. We use what you send to reply to you and for nothing else.
- Strict transport security, nosniff, frame-deny and a restrictive permissions policy, set at the edge.
- No cookies are set by this site, so there is no consent banner to dismiss.
- Fonts are self-hosted. Nothing on this page calls a domain that is not ours.
What we will send you.
Security review is a document exchange before it is a conversation. These go out on request, without an NDA for the first three.
- The architecture notes: trust boundaries, what talks to what, and where the dual-control checks sit in the call path.
- The invariant list: the named checks that run against a live book, and what each one would catch.
- A completed security questionnaire. Send yours and we will fill it in rather than returning our own format.
- Current certification and audit status, with dates. Ask and we will tell you exactly where each one stands rather than implying it from a badge.
Send the questionnaire.
We will fill it in honestly, including the rows where the answer is no, and you can decide from a complete picture rather than an optimistic one.